DBN+BPNN for fault classification

[SJHMAX]

I use DBN for the classification of bearing fault data sets at Case Western Reserve University. The data set includes 10 failures, but the final classification result is always a category. When you modify various parameters of the network in nnsetup.m, the result will change accordingly. For example, when you change nn.learningrate to 1, the classification will change from 1 to 4. Many people have encountered this problem, but No suitable solution was found. Sincere hope that someone can help me

[joey0111]

You can try to change the initial weight of the RBM and increase the number of training.

[SJHMAX]

@joey0111
thanks for your help.
i tried to set the initial weight to a random number between 0 and 1, and increased the number of training.The result is still the same as before.

[joey0111]

A random number between 0 and 1 is not enough. You can try multiplying all the initial weights by 0.01 or a smaller number. Also, have you normalized the input data?

[SJHMAX]

@joey0111
thanks....but it si still not working...and All data has been normalized.
The data source is the open bearing fault data set of Case Western Reserve University, USA, which is the original time domain signal. The training samples include 10 types of faults, each of which has 600 training samples, a total of 6000 groups, and 128 data points per group (all normalized).

[Passion-long]

Have you solved this problem? It seems that I have come across the same question.

[Yi0377]

i use the new matlab2018b。 | | Eric | | [email protected] | 签名由网易邮箱大师定制 On 4/26/2019 15:07，Yunlong Wang<[email protected]> wrote： Have you solved this problem? It seems that I have come across the same question. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub, or mute the thread.

[Passion-long]

I hava tried to use matlab2018b,but the result doesn't change at all.Is it possible that the code has a problem?

…

------------------ 原始邮件 ------------------ 发件人: "Yi Jing"<[email protected]>; 发送时间: 2019年4月26日(星期五) 下午3:44 收件人: "rasmusbergpalm/DeepLearnToolbox"<[email protected]>; 抄送: "王云龙"<[email protected]>; "Comment"<[email protected]>; 主题: Re: [rasmusbergpalm/DeepLearnToolbox] DBN+BPNN for faultclassification (#182) i use the new matlab2018b。 | | Eric | | [email protected] | 签名由网易邮箱大师定制 On 4/26/2019 15:07，Yunlong Wang<[email protected]> wrote： Have you solved this problem? It seems that I have come across the same question. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub, or mute the thread. — You are receiving this because you commented. Reply to this email directly, view it on GitHub, or mute the thread.

[SJHMAX]

I hava tried to use matlab2018b,but the result doesn't change at all.Is it possible that the code has a problem?
…
------------------ 原始邮件 ------------------ 发件人: "Yi Jing"[email protected]; 发送时间: 2019年4月26日(星期五) 下午3:44 收件人: "rasmusbergpalm/DeepLearnToolbox"[email protected]; 抄送: "王云龙"[email protected]; "Comment"[email protected]; 主题: Re: [rasmusbergpalm/DeepLearnToolbox] DBN+BPNN for faultclassification (#182) i use the new matlab2018b。 | | Eric | | [email protected] | 签名由网易邮箱大师定制 On 4/26/2019 15:07，Yunlong Wang[email protected] wrote： Have you solved this problem? It seems that I have come across the same question. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub, or mute the thread. — You are receiving this because you commented. Reply to this email directly, view it on GitHub, or mute the thread.

这个工具箱是没有问题的，就是不好调试，因为需要人工进行。

我用自己的数据，总共十类，用这个工具箱可以达到95的准确率。